The Kek C-Band Rf System for a Linear Collider H

TU201 Proceedings of LINAC 2004, Lübeck, Germany

THE KEK C-BAND RF SYSTEM FOR A LINEAR COLLIDER H. Matsumoto#, Shigeru Takeda, S. S. Win, M. Yoshida, KEK, Tsukuba Japan H. Baba, T. Shintake, SCSS Group, RIKEN, Harima Japan J. S. Oh, PAL/POSTECH, Pohang, Republic of Korea Y. Takasu, University of Tokyo, Tokyo Japan F. Furuta, University of Nagoya, Nagoya Japan

Abstract

C-band (5712-MHz) RF-system hardware R&D for an e+e- linear collider started in 1996 at KEK. We have al- ready developed three conventional 50-MW class klystrons, a smart modulator, and a novel HOM-free accelerator structure (Choke-mode type, full-scale high power model) [1], [2], [3], [4]. A very stable ceramic high voltage monitor was successfully tested up to 367-kV with 4.5-µsec pulses. Very good agreement in the expected division ratio and signal waveform fidelity was observed in high power tests. A new C-band SiC type high power rf-load, extending the power handling capability up to 50- MW is now being designed. It should have excellent mass production characteristics as it uses circularly symmetric TM011 chained cavities [5]. For the first ever, a high Figure 2: C-band main linac tunnels. The klystron gallery power prototype rf compressor (SLED III) cavity made of is 4.5-m in diameter and the linac tunnel is 3.0-m in di- a low thermal expansion material (Super Invar) was de- ameter. signed to provide stable operation even with a very high ear collider capable of undertaking the next essential Q of 200-k, it was operated up to a 135-MW peak output physics programs. Choosing a C-band technology entails power in 0.5-µsec rf pulses compressing input 45-MW a minimum of R&D thus facilitating early deployment 2.5-µsec pulses [6]. The C-band linac rf-system will be and reliable operation. The goal is to enable an early start used for production work in the SASE-FEL (Spring8 to the physics program, so as to be as concurrent as possi- Compact SASE Source, SCSS) project at SPring-8 [7]; ble with the LHC operation. Once a new particle thresh- SCSS will also serve to not only verify the design con- old is opened with LHC, all angles of the new physics cepts and components, but will also provide realistic regime can be thoroughly studied in the more straight- experience and lessons which can eventually be deployed forward clean experimental environment of e+e- colli- in the main linac rf system for a future large scale linear sions. collider. The main linac system is the heart of the linear collider. It is a huge system, composed of thousands of repetitions INTRODUCTION of common RF-units. Therefore, in order to realize a suc- cessful physics program, these RF-units have to meet The C-band main linac design and development has strict requirements for: (1) High reliability, (2) Simplicity, been motivated by the increasingly urgent need for a lin- (3) Easy operation, (4) Reasonable power efficiency, and (5) Low cost. These desiderata provide boundary conditions for our design work. Especially the first three items are crucial for such a large scale system to be operated at all; only after clearing them are the detailed dis- cussions about energy efficiency or upgradeability meaningful. As the C-band frequency is only two times higher than the S-band, the size of the accelerator structure is reduced by half. But more impor- tantly, currently available technology for fabricating the components can easily meet the accuracy required for the C-band, thus removing the risk of a non-manufacturable design. This single choice thus directly results in possibilities for high reliability, Figure 1: One unit of the C-band main linac. simplicity, and hence low cost.

#hiroshi.matsumoto@kek.jp

256 Technology, Components, Subsystems RF Structures Proceedings of LINAC 2004, Lübeck, Germany TU201

The total C-band main linac rf-system for a 500-GeV over long-term operation. Accord- C.M. energy includes about 2000 rf-units as shown in ingly we decided conservatively on Figure 1. In total about 8000 accelerating structures and a ceiling value of 300-400 about 4000 klystrons with modulators are needed for the Jules/pulse for the beam power in two main linacs. The numbers of each component are the klystron, while keeping the large, but still not enough to bring about the drastic cost maximum cathode emission load- reduction allowed by full mass-production. Therefore, ing to less than 10 A/cm2, and with from the start of the design of each component, maximum the maximum surface electrical efforts toward cost reduction for mid-scale production are gradient of the electrodes to less absolutely necessary. Accordingly the C-band group has than 22-kV/mm [4]. The experi- been inventing novel ideas and carrying out special cost mental test results of the three kly- reduction R&D. strons are summarized in Table 1. We propose that the C-band frequency allows the best The first tube (E3746-#1) em- set of trade-offs for meeting the demands. And the SCSS ployed a conventional design such project will give an opportunity for a realistic application Figure 3: as having only a single-gap output of the C-band rf technology. TOSHIBA-E3746. structure. Its main propose was to fix the mechanical dimensions of SYSTEM DESCRIPTION Table 1: The experimental result for C-band klystrons Each unit in the main linac rf-system is composed of E3746 No. 1 No. 2 No. 3 two 50-MW klystrons, their pulse modulators, one rf- Output power [MW] 50 (48) 54 55 pulse compressor, four 1.8-m-long choke-mode accelerat- Pulse width [µsec] 1 (2.5) 2.5 2.5 ing structures and an associated wave-guide-system as Repetition rate [pps] 50 (20) 50 50 shown in Fig. 1. The accelerating gradient is 35-MV/m Power efficiency [%] 42 44 45 under full beam loading. This was chosen as a practicable Output gap 1 31) 31) accelerating gradient after studying results from the S- 1) travelling-wave 3-gap output cavity. band frequency high gradient tests done between 1987 the electron gun, beam dump (collector), and the output rf and 1994 at KEK. windows; and also to characterize the C-band perform- The system will be installed in two concentric tunnels ance. The 2nd and 3rd klystrons were developed (in 1997 with circular cross section diameters of 3-m and 4.5-m for and 1998) to increase rf power efficiency, this was ac- the accelerator and klystron galleries, respectively. For complished by the newly introduced 3-cell traveling wave structural stability, the tunnels as shown in Figure 2 have output design. to be constructed in a very stable stratum such as granite. Power efficiency improved to 44% HARDWARE R&D RESULTS for the 2nd and to

We started hardware R&D in April 1996, and with the 45% for the 3rd exception of the high-power rf pulse compressor, by June klystron. Agree- 2003 we had developed most of the hardware components ment to within 1% and tested their performances. over the operating Figure 4: Typical efficiency and range was found Waveguide Components RF output power characteristics at between simulation saturation output rf power as a (FCI code) and a We chose a conventional rectangular EIA-WR187 function of beam voltage. precision calorimet- (47.55-mm x 22.15-mm) waveguide to make the system ric power measurement system. Figure 4 shows a typical simple. The rf transmission loss in the waveguide is – experimental result for the traveling-wave 3-gap klystron. 0.032 dB/m, which allows keeping the waveguide rf power loss budget to less than 5%. We have developed Modulator Power Supply various new waveguide components: among which there is a low cost yet highly reliable unisex type rectangular We focused our modulator R&D work on reducing the vacuum coupling flange (so-called MO type) [8]. It is fabrication cost and improving the reliability. As a first now commonly used in various facilities not only for C- step, we developed a prototype modulator, with features: band applications. (1) Direct HV charging from an inverter power supply, (2) No de’Q-ing circuit, (3) Much smaller in size than the Klystron R&D usual modulator, (4) Using existing low-risk reliable circuit components, such as the thyratron tube for the PFN We have successfully developed a 50-MW class sole- switching. noid focus type klystron (the TOSHIBA E3746 series), To reduce modulator size and permit removing the which meets the requirements for a 500-GeV linear col- de’Q-ing circuit from the PFN, we employed an inverter lider. type DC-HV power supply (the EMI-303L, U.S.A). A We have a clear design goal of ensuring high reliability first model was built in a compact metal cabinet with di-

Technology, Components, Subsystems 257 RF Structures TU201 Proceedings of LINAC 2004, Lübeck, Germany mensions 1.6-m (W) x 2-m (H) x 1.2-m (D). The fluctua- the TOSHIBA Co. in Japan and it was tested along with tion in the measured output voltage was measured to be the rest of the modulator beginning in March 2003 at less than ±0.17% (at 3σ), which meets the energy stability SPring-8. It generates a maximum output voltage of 50- requirement for the linear collider. The timing jitter and kV and provides an average power of 30-kW (or a peak drift of the pulse output is around 2-nsec (at 3σ) over a 4- of 37.5-kJ/sec); this supply can drive a 50-MW klystron hour run at 50-pps [9]. In 2003, this modulator concept at up to a 60-pps repetition rate delivering a 350-kV beam was accepted in voltage. As shown in Figure 6, we obtained an output China for the C-band Klystron voltage regulation of within ±0.1% with this first proto- Shanghai light type model. source. They fabri- RF Pulse Compressor cated it there them- 1 m 1.5 m selves, and it was At the present, initial tested in May 2004. Inverter HV 1 m testing of a high power rf The next step in power supply pulse compressor was be- the modulator de- gun at KEK at the end of velopment was to 2003. The prototype rf cav- install everything Modulator (Oil tank) ity uses a copper plated except for the in- Invar metal, this permits verting H.V. power Figure 5: A new developed oil- simplifying the temperature Figure 7: Typical pulse supply in an insu- filed modulator. control system for the rf compressor cavity rf power lating oil-filled metal compressor and thus con- waveforms. tank or cabinet of very tributes to reducing the cost of the total system [6]. Figure compact size as shown 7 show a very preliminary experimental result of a 135- in Figure 5 [10]. This is MW peak output power, 0.5-µsec pulse width at a 50-pps also very compact, being repetition rate with a total multiplication factor of 3.0. only 1.5-m (W), 1-m (H) From in this figure, we see the need to improve the flat- and 1-m (D). The proto- ness of the top of the output waveform, and also need to type was developed by increase the power gain factor from 3 to 3.3 for a realistic the NICHIKON Co in application. Japan. Testing was be- The thermal stability of these rf compressor cavities gun in March 2003 at Figure 6: Measured output provides an order of magnitude better performance than SPring-8. We obtained voltage stability at 50 kV and that of copper alone. No unusual vacuum out-gassing was the expected results as to 60 pps repetition rate. found even while in high power operation. pulse wave shape, volt- Unfortunately, after a high power test in March 2004, a age and flatness, and were able to verify the operational large water leak developed in one of the compressor cavi- repeatability of the unit and also its markedly reduced ties, which was flooded. This caused corrosion to start at EMI and noise generation. The main parameters for the the junction between the bare Invar and copper metal. modulator and new inverter power supply are listed in Table 2. RF Structure A new inverter H.V. power supply was developed by The C-band Choke-Mode type damped rf structure was Table 2: Main parameters of oil filed new modulator developed in 1998, and its performance has been con- Modulator peak output power: 111 MW firmed with the ASSET facility at SLAC [3]. A cut-away Average output power 46.7 kW view of the cavity is PFN charging voltage Nominal: 44 kV shown in Fig. 8. The Maximum: 50 kV dark square shapes (top Peak switching current: 5.4 kA and bottom) in Fig. 8 HV pulse width: 3.5 µsec show where the ring Pulse repetition rate (max): 60 pps shaped SiC HOM ab- Output voltage reputability: ±<0.5 % sorbers go in the as- Thyratron timing jitter: <5 nsec sembly. PFN impedance: 4.3 Ω PFN cell number: 18 Sections One particular advan- Transformer step-up ratio: 1:16 tage is that since all of Cabinet size (W) x (H) x (D): 1.5 x 1 x 1 m the parts are completely Inverter output voltage: 0 ~ 50 kV axially symmetric, they Output current (peak): 30 (37.5) kJ/sec can be machined on a Output voltage regulation: ±<0.07 % turning lathe, this easier Power factor at full load: >85 % machined type of cavity Power efficiency at full load: >85 % is very advantageous in Figure 8: A cut away view of the Cabinet size (W) x (H) x (D): 48 x 45 x 63 cm C-band choke-mode rf structure.

258 Technology, Components, Subsystems RF Structures Proceedings of LINAC 2004, Lübeck, Germany TU201 mass production. The first high power prototype model is primary field emission current from the observed dark being fabricated by the MITSUBISHI HEAVY current. The analysis shows that the primary field emis- IINDUSTRY Co. in Japan. The main parameters of the rf sion current from cathode surface is quite small for a Mo Table 3: Main parameters of Choke-mode rf structure surface, and the enhancement effect is small for a Ti sur- Frequency: 5712 MHz face. From this analysis, it is strongly suggested that Mo Phase-shift per cell: 3π/4 is most suitable material for the cathode and Ti for the Field distribution on the axis: Quasi-C.G anode. This was verified by experiment using Mo cathode Quality factor (average): 10300 and Ti anode electrodes; a field gradient of 130-MV/m Attenuation parameter: 0.53 was achieved with a total dark current below 1-nA when Filling time: 290 nsec the separation gap was 0.5-mm as shown in Figure 10. Shunt impedance (average): 58.5 MΩ/m Electric filed ratio of Es/Ea: 2.2 (max) Roller Cam Precise Active Mover Iris aperture up-stream: 17.330 mm down-stream: 13.587 mm The new roller cams mover unit is comprised of two Disk thickness: 4 mm roller cams, their stepping motors drivers, two linear slid- Number of cells: 91 ers and support Number of coupler: 2 frames as shown in (field symmetry & double irises) Fig. 11. We used Structure active length: 1.8 m 72-mm diameter structure are listed in Table 3. roller cams to pro- We decided to use a quasi-constant-gradient for the vide ±1.4-mm of electric field distribution along the axis of the structure; positioning area in this minimizes the surface electrical gradients, which con- the horizontal and tribute strongly to breakdown problems in high gradient vertical respec- operation. Doing this, we have successfully kept Es/Ea to tively. We do not a maximum of only 2.2. Higher trapped modes (HOM) of use V-blocks and small amplitude appeared at 20-, and 23-GHz in first pro- Figure 11: A cutaway drawing of flat plates fixed to totype rf structure. To eliminate them, the absorber disk the roller cams mover unit. 72-mm the structure (such thickness will be changed from 3- to 4-mm. Fig. 9 shows diameter roller cams give an ad- as is used in the a MAFIA simulation of the single bunch wake field. As justable range ±1.4-mm. magnet positioning can been seen, the wake field amplitudes are damped roller cams system enough before the arrival of the 2nd bunch, and also there in the SLAC FFTB). The expected maximum adjust- able range of ±1.4- mm was measured, and then the posi- tion repeatability Figure 12: Deviations of roller was tested and cams mover from reference. found to be less than ±0.1-µm anywhere within the positioning range Figure 9: A single wake field simulation of the C-band while loaded with a 50-kg dummy weight, shown in Fig- Choke-mode rf structure. ure 12. are no higher gap separation=0.5 mm 1500 trapped modes Mo-Ti Pulsed High-Voltage Monitor Mo found in the new Ti 115 1000 103 130 We have developed a very stable and accurate high- rf structure. 1nA voltage monitor, to be used for observing the klystron High Gradient pulse voltage [11]. Since it uses a ceramic material in a 500 capacitive type voltage divider (CVD), the capacitance A series of dark

DARK CURRENT [pA] division ratio can be kept quite stable even under tem- current measure- 80pA 0 perature changes, or changes in set-up configuration, or ments have been 50 100 150 changes in the mechanical stresses applied to the monitor FIELD GRADIENT [MV/m] made on two kinds port through the input lead. We successfully operated the Figure 10: Dark currents for Mo-Ti, Mo- of electrodes made monitor up to 367-kV, and 4.5-µsec pulses. The maxi- of Molybdenum Mo, and Ti-Ti electrodes as a function of field gradient at the cathode surface with mum voltage for the CVD test was limited by the avail- (Mo) and Tita- a gap separation of 0.5-mm. able modulator output voltage. nium (Ti). A new analysis method has been conceived of to separate the

Technology, Components, Subsystems 259 RF Structures TU201 Proceedings of LINAC 2004, Lübeck, Germany

A New C-band 50-MW SiC Type RF Load To saturate the FEL lasing in the 22.5-m long undulator There are no commercially available 50-MW class line, a low emittance beam current with as much as 2-kA vacuum rf loads for C-band frequencies. Therefore we peaks is required. The high peak current is generated by have constructed a new type rf load using SiC ceramic first compressing the bunch length in the injector and then indirectly cooled by water flowing through the structure; further in a magnetic-chicane bunch compressor. We will this upgraded the power handling density of the SiC ma- use four units of the 40-MV/m accelerating gradient C- terial from 100-W per cc to 300-W per cc, allowing a band accelerator, which should produce a beam energy much more compact overall size. The design uses a chain reaching 1-GeV with only a 30-m long accelerator; and of circular mode TM011 absorbers [5]. One particular ad- the shortest radiation wavelength should be 3.6-nm as vantage is that since the main parts are completely axially shown in Figure 13. symmetric, they can be machined on a turning lathe; thus this type of cavity has a big advantage in mass production REFERENCES because of its easier machining. [1] T. Shintake et al., “C-band Main Linac RF System for Linear Collider ”, KEK preprint 96-122, September. 1996. REALISTIC APPLICATION [2] H. Matsumoto et al, “Fabrication of the C-band (5712 MHz)

SCSS will provide one realistic application of one or a Choke-Mode Type Damped Accelerator Structure”, KEK preprint 98-143, September 1998. few rf-units in its linac. SCSS will be a soft X-ray SASE- [3] T. Shintake et al., “The first Wakefield Test on the C-band FEL machine aiming at demonstrating FEL operation Choke-Mode Accelerating Structure”, KEK preprint 99-11, below 10-nm wavelength with 1-GeV electron beam in May 1999. 2006~2007 [7]. The combination of a short period in- [4] H. Matsumoto et al., "DEVELOPMENT OF THE C-BAND vacuum type undulator and the high gradient C-band (5712 MHz) 50 MW CLASS PPM Klystron (II)", 26th Linear main accelerator makes the machine compact, enabling it Accelerator Meeting in Japan, Aug.01~03,2001, Tsukuba, to fit within a 100-m long tunnel. Japan The first project goal will be to generate 60-nm FEL [5] S. S. Win et. al., “A new C-Band 50 MW Class SiC RF load. from a 250-MeV energy beam by November 2005. II”, Proc. of 14th Symposium on Accelerator Science and Technology, November 11-13, 2003, Tsukuba Japan. Machine Configuration [6] M. Yoshida et al., “RF Pulse Compressor System”, 27th Li- nac meeting August 7-9, 2002, Kyoto. In the SCSS project, the following three key technolo- [7] T. Shintake et al., “Status of SCSS Project”, proc. of gies contribute to the compactness of the machine. (1) PAC2004, March 22-25, 2004, Gyeongju, KOREA. High gradient C-band accelerator. The accelerating gradi- [8] H. Matsumoto et al, “Development of C-band (5712 MHz) High Power Waveguide Components”, KEK preprint 97-50, ent can be as high as 40-MV/m, thus an accelerator only May 1997. 30-m long is enough to reach 1-GeV. (2) In-vacuum un-, [9] J-S. Oh et al, “Efficiency Analysis of the First 111-MW C- dulator, which enables creating a shorter period undulator band Klystron–Modulator System for Linear Collider”, KEK thus the required beam energy is lower, again reducing Preprint 98-32, March 1998. the accelerator size. It also contributes to shortening the [10]H. Matsumoto et al., “Closed type Compact Modulator for 50 FEL gain length. (3) Low emittance beam injector. The MW C-band Klystron”, Power Modulator Conference & short undulator period does require a low emittance elec- High Voltage Workshop, June 30 – July 3, 2002, Hollywood, CA. tron beam. We chose a HV (500-kV) pulse DC gun using a singlel [11]Y. Takasu et al., “Pulsed High Voltage Monitor Using Ce- ramic”, 27th Linac meeting August 7-9, 2002, Kyoto. crystal CeB6 thermionic cathode, which has the potentia [12]http://www-xfel.spring8.or.jp to generate a very small emittance beam while providing for a long lifetime.

Figure 13: Beam line layout in SCSS of 1-GeV case.

260 Technology, Components, Subsystems RF Structures